What it does
MeshServerWatch has two programs. An Agent runs as a service on each Windows or Linux server. A Manager that you run yourself receives the Agent reports. The Agent samples CPU every 30 seconds. It sends a report every 10 minutes by default, and the interval is configurable from 1 to 60 minutes. Each report covers CPU average and peak, memory used, each real disk with its percentage, load on Linux, uptime, OS name and agent version. Error-level operating system events are sent within about 30 seconds as one-line summaries. The Manager keeps 30 days of charts per server and applies thresholds that are set in the Manager. It raises an alert when a state changes. It raises an info-level cleared alert when the state returns to normal. The Manager has a browser console.
Capabilities
- The Manager is a single binary of about 8 MB that runs as a Windows service or a systemd service, with no runtime and no database server to install.
- Default thresholds are disk at 90% as a warning and 95% as critical, and CPU or memory averaging 90% or more for two reports in a row as a warning.
- A server is marked silent when no report arrives for 2.5 times the interval, and the message says the Manager only knows that nothing arrived, not that the server is down.
- Agents keep up to 6 hours of metrics and 500 alerts on disk while the Manager is unreachable, and send them in order when it is back.
- Agents are installed with the Manager's certificate fingerprint and refuse to send anything to a Manager whose certificate differs.
When to use it
- You want an email on Friday night when a disk fills up, rather than a ticket on Monday.
- You want server monitoring for Windows and Linux that runs on your own server, with no cloud account.
- You want 30 days of CPU, memory and disk charts for each server you monitor.
